#CriticGPT
🤖🔍 GPT-4 encontrando sus propios errores: llega CriticGPT

https://openai.com/index/finding-gpt4s-mistakes-with-gpt-4

#CriticGPT #OpenAI #IA #MachineLearning
May 30, 2026 at 11:24 PM
OpenAI has introduced CriticGPT, a tool designed to help developers identify and correct errors in code generated by ChatGPT. Based on GPT-4, CriticGPT aims to improve coding accuracy by catching subtle mistakes that may be difficult for AI trainers to detect.
July 8, 2024 at 1:12 PM
And then, there will be a chain of models CriticCriticGPT, CriticCriticCriticGPT and so on, each checking the previous. The chain will close on itself the last one being checked for lies by the ChatGPT itself.
---
Finding GPT-4’s mistakes with GPT-4 buff.ly/3zmrU3E
Finding GPT-4’s mistakes with GPT-4
CriticGPT, a model based on GPT-4, writes critiques of ChatGPT responses to help human trainers spot mistakes during RLHF
buff.ly
July 3, 2024 at 7:48 PM
It's a circular firing squad.

GPT-4 checking GPT-4. Currently focused on code, but likely extensible.

Oh, and love how they're telling on themselves with “Reinforcement Learning from Human Feedback” (RLHF). This is people. Training the model.

LLMs are STUPID. The ultimate mansplainers.
Finding GPT-4’s mistakes with GPT-4
CriticGPT, a model based on GPT-4, writes critiques of ChatGPT responses to help human trainers spot mistakes during RLHF
openai.com
July 3, 2024 at 2:33 PM
We just need CriticCriticGPT to fix CriticGPT
July 3, 2024 at 4:31 AM
I do not approve.
July 3, 2024 at 3:03 AM
OpenAI announces ChronosGPT: Intelligent AI Model sent back in time to stop errant sentient fork in CriticGPT no worries things should be back normal by two weeks ago
July 3, 2024 at 12:47 AM
Ask ChatGPT a question -> ChatGPT answers question -> CriticGPT checks the question -> ConfirmGPT checks CriticGPT's work -> CalcGPT recalculates all numbers (if necssary) -> CryptoGPT makes a crypto bet against being wrong -> CautiousGPT hedges that bet by buying gold -> CoalGPT burns some coal
Yo dawg, we heard you liked ChatGPT, so we put ChatGPT into your ChatGPT.
July 3, 2024 at 12:46 AM
This will only work if CriticGPT is AI in the same sense Amazon cashless stores are AI www.theguardian.com/commentisfre...
So, Amazon’s ‘AI-powered’ cashier-free shops use a lot of … humans. Here’s why that shouldn’t surprise you | James Bridle
This is how these bosses get rich: by hiding underpaid, unrecognised human work behind the trappings of technology, says the writer and artist James Bridle
www.theguardian.com
July 3, 2024 at 12:46 AM
July 3, 2024 at 12:34 AM
July 2, 2024 at 7:42 PM
Yes, let's ask the unreliable narrator which parts of the narration are unreliable 🙄 openai.com/index/findin...
Finding GPT-4’s mistakes with GPT-4
CriticGPT, a model based on GPT-4, writes critiques of ChatGPT responses to help human trainers spot mistakes during RLHF
openai.com
July 2, 2024 at 5:42 PM
NVIDIA CEO might be right about coding being dead because of AI — OpenAI's new CriticGPT model identifies ChatGPT's programming mistakes better than AI trainers
NVIDIA CEO might be right about coding being dead because of AI — OpenAI's new CriticGPT model identifies ChatGPT's programming mistakes better than AI trainers
OpenAI just unveiled CriticGPT to help identify and detect errors in code.
www.windowscentral.com
July 2, 2024 at 4:45 PM
It's "to assist trainers and reviewers identify bugs and errors generated by ChatGPT code. With companies wary of inaccurate chatbots, CriticGPT marks an attempt to earn confidence": OpenAI Releases CriticGPT to Catch Chatbot Hallucinations – Is It Enough? www.itprotoday.com/ai-machine-l...
OpenAI Releases CriticGPT to Catch Chatbot Hallucinations – Is It Enough?
After criticism of AI’s lying chatbots, OpenAI’s newest release, CriticGPT, is designed to point out ChatGPT’s errors.
www.itprotoday.com
July 2, 2024 at 4:20 PM
OpenAI has unveiled CriticGPT, a new AI model built on GPT-4, designed to enhance the quality of AI training. This model specifically critiques the responses from ChatGPT, aiding trainers in identifying and correcting errors during the model’s learning process.
July 1, 2024 at 10:04 PM
マトリョーシカ味ある。過渡期、良いね。「CriticGPTの出力は必ずしも正しいとは限らないが、人間のトレーナーがCriticGPTの支援を受けながら作業することで、間違いを見つけ出す精度を高められるという」。◆ OpenAI、ChatGPTの間違いを見つけるAI「CriticGPT」 - PC Watch buff.ly/3XS8z4m
OpenAI、ChatGPTの間違いを見つけるAI「CriticGPT」
OpenAIは27日、ChatGPTの間違いを見つけるためのAIモデル「CriticGPT」を発表した。ChatGPTと同様にGPT-4をベースに構築されている。
buff.ly
June 30, 2024 at 10:53 PM
CriticGPT: Новий ШІ від OpenAI, що викриває помилки ChatGPT і покращує точність кодування channeltech.space/ai/criticgpt...
CriticGPT: Новий ШІ від OpenAI, що викриває помилки ChatGPT і покращує точність кодування - Channel Tech
OpenAI планує інтегрувати моделі, подібні до CriticGPT, у свій процес маркування RLHF, надаючи своїм тренерам AI-допомогу.
channeltech.space
June 30, 2024 at 1:11 PM
<a href="https://www.itmedia.co.jp/news/articles/2406/28/news113.html" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link" target="_blank" rel="noopener" data-link="bsky">www.itmedia.co.jp/n...
OpenAI、GPT-4のエラーを指摘するGPT-4ベースの「CriticGPT」リリース
OpenAI、GPT-4のエラーを指摘するGPT-4ベースの「CriticGPT」リリース
OpenAIは、ChatGPTが生成するコードに潜むエラーを検出するモデル「CriticGPT」を発表した。人間がChatGPTの誤りを発見するのを支援する。
www.itmedia.co.jp
June 30, 2024 at 12:25 PM
AI is evolving, but accuracy is key. State-of-the-art models are like toddlers—brilliant but sometimes a hot mess. How do we keep them in check? 🧠

Source: MarkTechPost
search.app/WjPi9aFqDWJH...
OpenAI Introduces CriticGPT: A New Artificial Intelligence AI Model based on GPT-4 to Catch Errors in ChatGPT's Code Output
OpenAI Introduces CriticGPT: A New Artificial Intelligence AI Model based on GPT-4 to Catch Errors in ChatGPT's Code Output
search.app
June 30, 2024 at 2:34 AM
<a href="https://www.itmedia.co.jp/news/articles/2406/28/news113.html" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link" target="_blank" rel="noopener" data-link="bsky">www.itmedia.co.jp/n...
OpenAI、GPT-4のエラーを指摘するGPT-4ベースの「CriticGPT」リリース
OpenAI、GPT-4のエラーを指摘するGPT-4ベースの「CriticGPT」リリース
OpenAIは、ChatGPTが生成するコードに潜むエラーを検出するモデル「CriticGPT」を発表した。人間がChatGPTの誤りを発見するのを支援する。
www.itmedia.co.jp
June 29, 2024 at 11:34 PM